Confirmatory and exploratory analysis applied to pharmaco-EEG and related study data: contradiction or useful...

Besides hypothesis testing, which should be done as sparingly as possible, the measured or observed data should be described as extensively as possible. The traditional reliance on profiles of the mean responses may neglect useful information, and such profiles may also be misleading. With the aid of exploratory data analysis, different aspects of the structure of a data set can be considered. 'Data snooping' may discover coherences, non-trivial structures and peculiarities, which lead to a new hypothesis or to new mathematical-statistical models. It is, in our opinion, a necessity to consider exploratory and confirmatory data analyses in conjunction. This will be illustrated by examples taken from pharmaco-EEG studies.